WHAT IS THE ROLE OF OPENNESS FOR CHINA’S ENVIRONMENT? AN

Transcription

Fifth Chinese Economy Conference of CERDI (20-21 Oct, 2005)
halshs-00570491, version 1 - 28 Feb 2011
WHAT IS THE ROLE OF OPENNESS FOR CHINA’S ENVIRONMENT? AN
ANALYSIS BASED ON DIVISIA DECOMPOSITION METHOD FROM THE
REGIONAL ANGLE
Jie Heψ
Observing the weakness in the previous structural analyses on EKC formation,
in this paper, author deepens the analysis into the detailed data of production and
SO2 emission intensity of China’s 29 industrial sectors (occupying over 98% of the
total industrial production) in each province during 1991-2001. With the aid of
Divisia Index Decomposition method, the variation of the provincial-level industrial
SO2 emission with regard to the original level of 1990 is decomposed into the
contribution from its three determinants: the variations in its production scale, its
composition transformation and its technique character changes. The following
analysis aims to reveal regional differences in environmental impact of
industrialization and to further interrogate the potential links between these regionspecific environmental impacts of industrialization and development of commercial
openness in each province.
Key Words: Openness, pollution, China, Decomposition, region.
ψ
CERDI, Université d’Auvergne, France. [email protected]
1. Introduction
Most of the literatures aiming at explaining the formation of the Environmental Kuznets
Curve in an economy generally agree with the Grossman (1995) decomposition and
distinguish the determination role played by its scale, composition (structure) and technique
characters. Although originally the composition effect is to describe the actual pollution
performance of the economy’s industrial structure, constrained by data availability, most of
existing literature on this topic employed the capitalistic ratio (or the capital abundance
degree) calculated as the total capital employed in production divided by total labor (K/L). As
generally (or traditionally) polluting heavy industries employ more capital in their production,
the common sense suggests that we can use this capitalistic ratio as an approximation for
environmental performance of the industrial structure of an economy and suppose higher
capitalistic ratio means more intensive polluting industry concentrated in economy. However,
the rise of the knowledge and high-tech industries during the last several decades obviously
disrupts this assumption, since the production process of most of the capital-intensive new
high-tech industries, as the information industry, are not polluting at all.
Observing the weakness in the measurement of composition effect, in this paper, author
deepens the analysis into the detailed data of production and SO2 emission intensity of
China’s 29 industrial sectors (occupying over 98% of the total industrial production) in each
province during 1991-2001. With the aid of Divisia Index Decomposition method, the
variation of the provincial-level industrial SO2 emission with regard to the original level of
1990 is decomposed into the contribution from its three determinants: the variations in its
production scale, its composition transformation and its technique character changes. The
following analysis aims to reveal regional differences in environmental impact of
industrialization and to further interrogate the potential links between these region-specific
environmental impacts of industrialization and development of commercial openness in each
province.
The organization of the paper is as following, in section 2 we introduce the Grossman
Decomposition and then we show in section 3 how the Grossman decomposition can be
related to the Divisia index decomposition method and how we need to adapt the Divisia
index decomposition method to China’s data. The decomposition results that distributed the
variations of China’s industrial SO2 emission into the contribution from scale enlargement,
structural transformation and technical progress during 1991-2001 are then discussed in
section 4. Following in section 5, we employ the decomposed contribution of the three
structural determinants in emission variation to investigate the potential impact of
international trade on industrial SO2 emission via these three determinants. Finally, we
conclude in section 6.
2. Grossman Decomposition
Most theoretical explanations for EKC formation base their reasoning on the demandside factor. These analyses generally consider the decoupling of economic growth and
environmental deterioration as the result of the trade-off between utility from normal good
consumption and disutility caused by pollution. However, demand-side causes are only the
necessary conditions for the appearance of EKC turning point. Without satisfying certain
supply-side conditions, such as technological capacities in pollution abatement or production
process and the de-pollution trends in industrial structure transformation, a simple rise of
public demand for a better-environment may only lead economic growth to a halt and cannot
realize a win-win result. As this paper is interested in industrial SO2 emission situation in
China, a developing country, whose most important objective is still to enrich its population
at the moment, the analyses on the supply-side emission determination factors are actually
more appropriate.
More or less, based on macroeconomic foundation, some theoretical analyses have
already involved in their discussion the supply-sides pollution determinants and explored the
structural explanation for the formation of EKC. The most widely accepted supply-side
pollution determinants are those proposed by Grossman (1995), which regarded pollution as a
by-product of production activities. According to him, the final total emission results can be
calculated by equation (1).
n
Et =Yt ∑I j,t S j,t
(1)
j =1
Where Et means the emission in year t. j=1,2,…,n represents the different sectors in the
economy, Yt is the total GDP in year t, it can also be presented by the sum of value added of
the n sectors, so Yt=ΣYj,t. Ij,t is the emission intensity, the average quantity of pollution
emitted for each unit of product in sector j, so Ij,t=Ej,t/Yj,;t. Sj,t=Yj,t/Yt, it represents the ratio of
the value added of sector j in total GDP. According to this equation, total emission can be
considered as product of the total value added of economy Yt and the average sectoral
emission intensity weighted by the ratio of each sector’s GDP in total economy Σj(Ij,tSj,t).
If we make total differentiation with respect to time and then divide the whole equation
by total emission results, E, we reproduce the Grossman decomposition in equation (2).
^
^
^
^
E =Y +∑e j S j +∑e j I j
j
(2)
j
Where ej=Ej/E represents the ratio of emission from sector j in total economy and
^
X =(dX / dt)/ X t ,X ∈{E,I,S,Y} . This decomposition defines the famous three determinants of
emission. Y^ denotes the scale effect, which is expected to be a pollution-increasing factor.
“All else equal, an increase in output means an equiproportionate increase in pollution”. The
^
composition effect is represented by S j , the changes in Sj over time represent the influence on
emission of a change in the composition of economic activities. All else equal, if the sectors
with high emission intensities grow faster than sectors with low emission intensities, the
composition changes will result in a upwards pressure on emission, so that total emission will
^
grow at a faster rate than income. I j represents the technique effect. The decrease in sector
emission intensities, as the results of the use of more efficient production and abatement
technologies, can reduce emission increasing pressure given the same quantity of economic
growth and fixed industrial composition mix.
Grossman decomposition is actually based on the decomposition principal of Divisia, in
the following sector, we will explain in details about the decomposition steps.
3. Linkage between the Grossman decomposition to Divisia index decomposition
method
3.1. The origin of Divisia index
The Divisia index is a continuous time index number formula due to François Divisia
(1925). This index has already been widely used in theoretical discussions of data aggregation
and the measurement of technical change. It is defined with respect to the time paths of set
variables. For example, at given period t, the prices [P1(t), P2(t), P3(t),…, PN(t)] and
commodities [X1(t), X2(t), X3(t),…, XN(t)]. Total expenditure on this group of commodities is
given by:
Y(t)=P1(t)X1(t)+P2(t)X2(t)+P3(t)X3(t)+…+PN(t)XN(t).
(3)
Using dots over variables to indicate derivatives with respect to time, the total
differentiation of (3) yields:
•
•
•
Y( t) N Pi(t) X i(t ) Pi( t) N Pi(t) X i(t ) X i( t) .
=∑
+∑
Y( t) i = 1 Y (t ) Pi( t) i = 1 Y (t ) X i( t)
(4)
The first summation of the right-hand side of (4) defines the Divisia index of prices and
the second defines the Divisia quantity index. Both indexes are weighted averages of the
growth rates of the individual Pi(t) and Xi(t), where the weights are the components’ shares in
total expenditures:
Pi(t )X i( t)
.The sum of equation (4) thus defines the rate of change of the
Y (t)
aggregate price and quantity indexes. The levels of these indexes are obtainable by line
integration over the trajectory followed by the individual prices and quantities over the time
interval [0,T]. For the quantity index, the line integral has the following form:1
[
I q(0,T)=exp{ ∫
N
•
∑ Pi(Yt)(Xt)i(t) XX ii((tt))
i= 1
] }.
(5)
We note that the Divisia index is defined using time as a continuous variable. This is
actually not appropriate for empirical analysis, where the data on both price and quantity
typically refer to discrete points in time.
To resolve this problem, we need to approximate the continuous variables of equation
(4) with their discrete time counterparts. The approach of Törnqvist (1936) suggests
approximating the growth rate of prices and quantities by logarithmic differences and the
continuous weights by two periods arithmetic averages. So the Törnqvist approximation to the
growth rate of the Divisia quantity index can then be written as:
N
∑0.5[ P YX
i= 1
i ,t
i ,t
t
+
Pi ,t-1 X i ,t -1
X i ,t
Yt -1 ]ln( X i,t -1 )
(6)
Equally, the Divisia price index can be calculated as:
N
∑0.5[ P YX
i= 1
i ,t
t
i ,t
+
Pi ,t -1 X i,t -1
Pi,t
Yt -1 ]ln( Pi ,t -1 )
(7)
Comparing equation (6) and (7) with Grossman decomposition equation (2), we can see
that Grossman decomposition is exactly following the decomposition principle of Divisia to
decompose the multiplication between composition and technique effect into two independent
terms.
3.2 Adaptation of the Divisia index decomposition method to China’s provincial
level panel data
If we agree with Grossman (1995) and consider emission as a by-product of production,
Based on the decomposition idea of Grossman (1995), the industrial SO2 emission in each
1
For more discussion of Divisia integrals, see Richter (1966) and Hulten (1973).
province i during period t can be writen as a product of its three determinants as in equation
(8).
SO2 ,it = Yit × ∑(
scale
j
Y j,it
Yit

×
SO2j,it
)
Y
j,it



.
Compositio n Technique
(8)
The index j=1, …, n signifies different industrial sectors. Yit presents the total industrial
GDP in province i during period t and SO2,it signifies the total industrial SO2 emission of the
same province. Yj,it means the GDP created in industrial sector j of province i during period t
and SO2j,it is the corresponding sector-level SO2 emission. Corresponding to equations (1) and
(2), (Yj,it/Yit) gives the proportion of product of sector j in total industrial product of province I
and the term (SO2j,it/Yj,it) calculates the sector-specific pollution intensity. In this equation, the
term (Yit) corresponds to scale effect, the term (Yj,it/Yit) is the composition effect, and finally
the term showing the emission intensity detailed in provincial and sectoral level represents the
technique effect.
According to equation (8), we see that to make a thorough decomposition of the
provincial level SO2 emission into the three effects, we need to possess the detailed
information on production and emission intensity on both provincial and province-specific
sectoral level. Although Chinese Industrial Economic Statistic Yearbook and China Energy
Databook 6.0 only provide us with detailed sectoral production mix for 13 industrial sectors
on both national and provincial level for the period of 1990s, we do not possess the related
emission effluent rate for the province-specific industrial sectors during the same period.2
Given the current data condition, a provincial level complete Divisia index decomposition
analysis is actually not feasible. But fortunately, the national-wide average emission
intensities for different industrial sectors are reported annually since 1991, with the help of
this data series, it is still possible for us to carry out a relatively “incomplete” decomposition
analysis to investigate the role of scale enlargement, product structure changes and technical
progress in industrial SO2 emission variation on both national and provincial levels.
To adapt the decomposition method to our data situation, we need to make following
adaptation in the original Divisia decomposition method.
2
The 13 industrial sectors are total mining industry, food and beverage, textile, paper, total power industry,
chemical materials, pharmacy, fiber, non-metal products, metal processing and smelting, metal products,
machinery and the other industry, these data comes from China Industrial Economic Statistic Yearbook (19892002), and their total value added accounts up to 98% of the total industrial VA in each province. Therefore we
are quite confident on the representation capacity of this indicator.
Firstly, we can express provincial and national average level emission intensity in sector
SO2
SO2
j as I jt = Y jt jt and I jit = Y jitjit , respectively. For the proportion of sector j in total industrial
Y jit
product, its national and provincial level situation is represented by S jt = YYjtt and S jit = Yit .
Due to the data constraint, we only have national level average sectoral emission
intensity Ijt. Therefore, we need an efficiency deflator ejit for each sector j of province i to
construct the following identity equation.
n
n
Y jit SO2jit
×
)=Yit × ∑(S jit ×I jt ×e jit ) ,
Yit
Y jit
j =1
j =1
SO2 it =Yit × ∑ (
SO2jit Y jit
with e jit =
.
SO2jt / Y jt
(9)
For simplification, we suppose all the industrial sectors in the same province share the
same emission intensity deflator eit, that means
n
∑(S jit × I jit )
j =1
eit = n
∑(S jit × I jt )
j =1
=
SO2it
n
Yit × ∑(S jit × I jt )
.
(10)
j= 1
Replacing eit into (9) and combining (10), we have a new expression for the industrial
SO2 emission in province i during period t as
n
SO2 it =Yit × eit × ∑(S jit × I jt ) .
(11)
j =1
Therefore its fractional variation during period t can be written as,
n
∑(S jit × I jt )
SO2it = Yit × eit × j =1
SO2it -1 Yit -1 eit -1 n
∑(S jit -1× I jt -1 )
.
(12)
j=1
Applying Divisia index decomposition method introduced in last section to equation
(12) and following Törnqvist continuous hypothesis, we have,
n
SO2it
= Yit × eit × exp( ∑
SO2it -1 Yit -1 eit-1
j =1
S jit ×I jt ×ln(
n
S jit
)
S jit -1
∑ (S kit ×I kt )
k =1
I jt
n S jit × I jt ×ln( I jt-1 )
)× exp( ∑
).
n
j =1
∑ (S kit ×I kt )
k =1
(13)
Transforming the rolling base year specification in (13) into fixed base year
specification with the reference as the original year 0 and changing the equation (14) into
logarithmic form, we have,3
3
Please note that
SO2 jt = SO2 jt × SO2 jt −1 ×⋅⋅⋅⋅⋅⋅× SO2 j1
.
SO2 j0 SO2 jt − 1 SO2 jt −2
SO2 j 0
ln(
T
n
T
n
SO2 it
S jit × I jt
d ln(S jit)
S jit × I jt
d ln(I jt )
)= ln( Yit )+ ln( eit )+ ∫ ∑ ( n
×
×dt ) + ∫ ∑ ( n
×
×dt)
SO2 i0
Yi0
ei0 t =1 j= 1
dt
dt
t =1 j =1 ∑
∑ ( S kit ×I kt )
(S kit × I kt )
k =1
(14).
k =1
A more applicable equation that we will use in decomposition step is
ln(
T n
T n
SO2 it
S jit
I jt
)= ln( Yit )+ ln( eit )+ ∑ ∑ [(w jit + α(w jit - w jit -1 ))×ln(
)] + ∑ ∑ [(w jit + β( w jit - w jit -1 ))×ln(
)]
SO2 i0
Yi0
ei0 t =1 j =1
S jit -1
I
jt -1
t =1 j= 1
+residualsi.
Where
w jit = n
(15)
S jit × Iit
∑(S kit × I kt ) . We equally suppose α=β=0.5.
k =1
We call the first terms ln( YYiit0 ) on the right-hand side of the equation (4.19) as scale
T n
S jit
effect, the term ∑∑ [ (w jit + α(w jit - w jit-1 ))×ln( S jit-1 )] as structure effect and the sum of ln( eeiit0 )
t =1 j =1
T n
I jt
and ∑ ∑ [( w jit + α(w jit - w jit -1 ))×ln( I jt -1 )] as technique effect.
t =1 j=1
3.3 efficiency discussion—the adapted Divisia decomposition method vs. the already
existing decomposition analyses
In most of the traditional Divisia decomposition analyses that are applied to emission
studies, the authors often choose to deepen their starting point from the energy input mix level
of each industrial sectors, such as OECD (1991) or Torvanger (1991). This depth, however, is
not possible given our currently data situation. However, to explore the very detail of each
possible emission sources, these authors are also obliged to follow the the “bottom-up” logic,
according to which the final total emission was often an estimated value obtained by
multiplying the consumption of individual energies by their corresponding emission
coefficients and then summing them up. The emission coefficients that they used were
generally directly borrowed from some estimation based on developed countries’ experiences.
However, considering the potential purity divergence between the energies in different
countries that are officially categorized into the same type and the potential emission effluent
rate differences related to different combustion technologies, we doubt about the pertinence to
directly extrapolate one country’s emission result according to the emission effluent rate
estimated from another one.
In this paper, as our aim is to get a better understanding on the determination
mechanism of the variation of total industrial SO2 emission, we choose to follow the “topdown” logic and carry out a direct decomposition analysis starting from the actually reported
total industrial SO2 emission statistics. One of the most important advantages of our “topdown”-style decomposition method is, by simply striping the scale and composition effect
from the actually happened total emission variation, it actually guarantees us with a more
general but also more credible technique effect.
4. Decomposition results
Table 1 displays the total variation of industrial SO2 emission in year 2001 with respect
to that of 1991 and the contributions from the three effects computed according to the
equation (15). Although there are always some unexplainable residuals, by employing the
Adjusted Weighted Decomposition principle combined with the rolling base year index
decomposition method, we actually get relatively small residuals.
Table 1 provides the decomposition results expressed in both total and annual average
percentage change forms. It seems during the last ten years, On national level, the total
industrial SO2 emission in year 2001 has increased by 30%. industrial SO2 emission generally
experienced an increasing tendency in most of the provinces, We observe equally the big
regional differences in SO2 emission’s variations, especially the annual average percentage
changes. With an annual growth rate of 2.7%, the annual variations of some provinces are up
to over 5% (Guangdong, Hainan, Henan and Shanxi). While during the same period, some
other provinces, such as Beijing, Liaoning, Heilongjiang, Shanghai and Jianxi, has
experienced net environmental amelioration.
The decomposition result revealing the most important contributor in industrial SO2
emission is industrial production scale enlargement. Varying from 11% to almost 22%, the
industrial SO2 emission increase caused by scale effect seems to follow remarkable average
percentage changes each year in all the provinces.
The decomposition reveals the expansion of the industrial production scale as the most
important factors to explain of the rapid increase of industrial SO2 emission. However,
compared to the actual emission increase, the production scale expansion realized during the
last ten years is obviously much more important, especially for some coastal and frontier
provinces. (Cf. Figure 1). It is hence from the sides of composition and technique effects that
we should look for the factors that help in reducing the potential emission increasing related
to expansion of scale effect.
Table 1. Total and annual average changes in industrial SO2 emission and the decomposed contributions from different structural determinants
Changes (2001 vs. 1991, 1991=1)
Annual average percentage changes (percent per year)
SO2
Scale
Composition
Technique
Residual
SO2
Scale
Composition
Technique
CHINA
130.439
256.639
119.618
-245.041
-0.777
2.693
9.883
1.807
-9.376
BEIJING
49.162
232.090
77.310
-259.323
-0.915
-6.854
8.784
-2.541
-9.998
TIANJIN
128.040
267.690
80.557
-216.435
-3.773
2.503
10.348
-2.139
-8.027
HEBEI
139.272
271.674
119.367
-250.521
-1.248
3.368
10.511
1.786
-9.619
SHANXI
157.026
225.890
148.945
-219.026
1.217
4.616
8.490
4.064
-8.156
INNER MONGOLIA
109.759
232.172
147.057
-268.752
-0.717
0.936
8.788
3.932
-10.391
LIAONING
72.482
214.766
118.545
-260.126
-0.703
-3.167
7.944
1.716
-10.032
JILIN
111.452
237.371
107.724
-233.034
-0.609
1.090
9.029
0.747
-8.828
HEILONGJIANG
99.522
233.792
103.433
-237.982
0.279
-0.048
8.864
0.338
-9.057
SHANGAI
87.504
247.627
118.522
-278.054
-0.592
-1.326
9.491
1.714
-10.768
JIANGSU
121.179
269.686
113.497
-260.443
-1.561
1.940
10.430
1.274
-10.045
ZHEJIANG
149.130
279.017
141.876
-270.699
-1.064
4.077
10.806
3.560
-10.471
ANHUI
108.601
239.398
124.139
-254.278
-0.658
0.829
9.122
2.186
-9.782
FUJIAN
152.592
291.408
139.658
-277.021
-1.454
4.317
11.288
3.397
-10.726
JIANGXI
92.205
212.095
144.209
-262.994
-1.107
-0.808
7.809
3.729
-10.153
SHANDONG
111.157
286.172
102.656
-276.636
-1.035
1.063
11.087
0.262
-10.711
HENAN
161.708
269.904
138.154
-245.803
-0.547
4.924
10.439
3.285
-9.410
HUBEI
109.608
249.096
97.816
-235.975
-1.329
0.922
9.556
-0.221
-8.965
HUNAN
116.016
216.396
124.157
-224.474
-0.063
1.497
8.025
2.187
-8.422
GUANGDONG
177.623
296.831
135.925
-253.895
-1.238
5.913
11.494
3.117
-9.765
GUANGXI
126.180
220.430
132.304
-225.738
-0.817
2.353
8.225
2.839
-8.483
HAINAN
165.664
263.425
143.381
-241.161
0.020
5.177
10.171
3.669
-9.202
SICHUAN
152.870
225.809
120.213
-193.033
-0.119
4.336
8.486
1.858
-7.545
GUIZHOU
138.195
216.258
159.329
-237.043
-0.348
3.288
8.018
4.768
-9.014
YUNNAN
138.670
241.519
134.817
-236.947
-0.718
3.323
9.218
3.033
-9.010
SHAANXI
106.130
232.684
135.682
-261.874
-0.361
0.597
8.812
3.098
-10.106
GANSU
111.893
220.307
131.854
-240.117
-0.151
1.130
8.219
2.804
-9.155
QINGHAI
116.915
245.187
151.676
-280.407
0.458
1.575
9.383
4.254
-10.861
NINGXIA
134.111
251.767
149.474
-268.412
1.281
2.978
9.673
4.101
-10.377
XINJIANG
116.626
273.549
137.692
-294.298
-0.317
1.550
10.587
3.250
-11.398
Note: Total change of SO2 is equal to the sum of the contribution from the changes in scale, composition and technique effect and the residual. Note that the residuals obtained
from the AWD rolling based year decomposition method are generally very small.
Region
<138.5%
138.5%159.5%
159.5%1810.5%
<01%
01%-22.5%
22.5%-43.5%
>43.5%
Scale effect
Composition effect
<0%
0%-2%
2%-4%
>4%
<-14.110.5%
-14.10.15% -13.69.5%
- 13.69.5% -13.28.5%
>-8.5%14.1%
Technique effect
Emission changes
Figure 1. Divisia Decomposition results
(Annual average percentage changes)
The general positive numbers associated with the composition effect reveals the second
pollution-increasing factor for most of Chinese provinces—a not surprising finding for a
country in its early phase of industrialization process. There are only three exceptions, Beijing,
Tianjin and Hubei, whose total composition effect appeared to be emission-reducing factor
and also several other provinces whose total contribution from composition effect to emission
increase is relatively small (below 2% per year), such as Shangdong, Jiangsu, Shanghai,
Liaoning and Hebei. A common point between these provinces are their geographical
location: they are all located in the east coast of China and their prosperous economic growth
are all benefiting from high degree of commercial and capital openness. We suspect the
emission reduction contribution from composition effect in these five provinces is actually
due to their participation into global production division system, in which, China, abundantly
endowed with cheap labor forces, possesses actually comparative advantages in the relatively
less polluting industries as textile, electronic equipment, etc. This idea can be better
represented by the panel of composition effect in Figure 1, in which the pollution increase
caused by composition effect is gradually reinforced when we move from east costal
provinces to those located in the central and west China, especially for these northern
provinces.
The regional disparity in the contribution of technique effect in emission evolves almost
in the same proportion as that of scale effect. Except that this time, we observe a factor
contributing to the reduction of emission. From Figure 1, we can see that it is the richer
coastal provinces generally possess a higher technique effect. This result actually confirms our
intuition that the technique effect increases with the income growth.
Adding up the three decomposed effects gives us the final emission variations. Clearly,
the variation of the emission finally observed in the reality is actually a specific result for each
province, whose formation actually depends closely on the characteristics of each province in
the aspects of production scale, economic structure and technical progress. Looking carefully
the six provinces whose annual average emission variation surpasses 4% in Figure 1, for the
three provinces located in coastal area, Zhejiang, Fujian and Guangdong, their pollution
situation is actually due to the fact that their strong expansion of industrial scale cancel out
totally the emission reducing contribution from the technique effect. While for the 3 inland
provinces, Sichuan, Shanxi and Henan, their emission increase is owing to the stronger
pollution increasing contribution from composition transformation with respect to their
relatively weaker technique capacity.
Figure 2 plots the detailed annual dynamic evolution of the national level industrial SO2
emission and its three contribution factors. Similar to the conclusion in Selden et al. (1999)
and Bruvoll and Medin (2001), our decomposition results also emphasize the technique effect
as the most important SO2 emission reducing factor. However, as there exists a mutual
cancellation effect between scale effect which creates emission and technique effect that
reduces the emission, it is actually the composition effect, whose contribution in emission
stays relatively small and the result of the mutual cancellation between scale and composition
effect that determine the final variation of the industrial SO2 emission.
Figure 4.2. Decomposition result on national level (1991-2001)
300
250
Index (1991=100)
200
emi
150
Scale
100
Composition
technique
50
00
01
20
20
98
99
19
19
97
19
95
96
19
19
93
92
94
19
19
19
-50
19
91
0
-100
Year
Figure 2. Decomposition results on national level (1991-2001)
5. Tracing the trade-related scale, composition and technique effect—an empirical
analysis based on the Divisia decomposition results
With the decomposed results of the contribution of scale, composition and technique
effects in the total industrial SO2 emission, we can now check how international trade affect
the final emission results in China.
5.1. Normalization of the variation index into the actual emission changes in quantity
From eq. (15), we saw the Divisia index decomposition methods actually reasons on
percentage change basis with reference to the original industrial emission level. Its
contribution is to decompose the percentage change in industrial SO2 emission, ln( SO2i,t SO2i,0 )
, into the contribution from scale enlargement ln( YYiit0 ) , composition transformation
T n
∑∑ [ (w jit + α(w jit - w jit-1))×ln( SSjitjit-1 )]
t =1 j =1
and
technique
progress
ln( eit ) +
ei0
T n
∑ ∑ [( w jit + α(w jit - w jit -1 ))×ln( IIjtjt-1 )] .
t =1 j=1
However, our econometrical analysis actually requires the scale, composition and
technique effect to be expressed in quantity of emission changes instead of the percentage
changes. Therefore, as the first step, we need to normalize the decomposition results for the
three effects. The normalization process is illustrated in equation (16)-(18). In this step, we
simply use the total SO2 emission variation in each year with respect to the reference original
year 0 as the base value and attribute this value to the three effects according to their
contribution shares.
ΔSO2,i,t,scale =
ln( Yi,t Yi,0 )
×(SO2
ln(SO2 i,t / SO2 i,0 )
T
Δ SO2,i,t,compositio n =
Δ SO2,i,t,technique =
(16)
i,t - SO2 i,0 )
n
∑ ∑ [(w jit +α(w jit - w jit -1 ))×ln( SSjitjit-1 )]
t =1 j =1
×(SO2 i,t - SO2 i,0 )
T n
I jt
ln( eit )+ ∑ ∑ [( w jit + α(w jit - w jit-1 ))×ln(
)]
ei0 t =1 j= 1
I jt -1
×( SO2 i,t - SO2 i,0 )
(17)
(18)
5.2. Econometrical strategy and estimation results
Our empirical strategy in this section is relatively simple. As the variations in industrial
SO2 emission related to international trade can be considered as a final combination of the
three partial changes caused by scale, composition and technique effects, the objective of this
section is to trace the indirect environmental role of trade through its separate impacts on the
emission variations that are contributed by scale enlargement, composition transformation and
technical progress.
Our investigation is carried out in two steps. In the first step, we regress the reducedform estimation function, which relates trade intensity ratio directly to each of the three partial
emission changes. In the second step, we regress the partial emission changes directly on their
corresponding effects, measured respectively by industrial GDP, capital/labor abundance ratio
and per capita GDP, by the Two Stage Least Square (2SLS) estimator. In this step, trade’s
impact on emission via the three effects is actually captured by the first stage of the 2SLS
estimation, in which we use trade liberalization degree as instruments for the scale,
composition and technique effects. Although in the second step, the three effects are not the
endogenous variables with respect to their corresponding emission changes, the 2SLS
estimator actually enables us to separately analyze the impact of trade on the scale,
composition and technique effect and how the three effects, in their turns, affect the total
emission results. Multiplying the coefficients of the structural determinants obtained in the
second-stage 2SLS estimation with the trade’s coefficient found by the first stage
instrumentation estimation, we can give a more direct description on how the trade intensity
variation affects emission through each of the three channels.
Detailed reduced-form estimation results for the three effects are reported in table 2, 3
and 4. Tables 5-8 list the 2SLS structural estimation results.
As a general character often observed in Asian countries’ industrialization histories is
that they often use foreign exchanges obtained from export to finance the import of machinery
and equipment that used to support the development of some strategic heavy industries. If
their export growth is stimulated by the demand from the world market, their import activity
is more policy-oriented. Considering the possible difference in the role of export and import in
economy and pollution, in the estimation carried out in this section, besides the (X+M)/GDP,
we also check the emission determination role of export and import separately.4
We firstly include the export and import intensity (X/GDP and M/GDP) separately in
the emission determination model B. Suspecting it might be the stock of the imported
equipment and machinery instead of the annual import flow that has the capacity to affect the
economic structure in China, we further replace in model C the import intensity (M/GDP) by
the variable Mstock, which measures the employment of the stock of the imported equipment
and machinery in total economy.5 This openness measurement is inspired by the CGE
production model of de Melo and Robinson (1990), in which the positive externality of the
international trade on developing economy is describe through both the increasing export
activities and the accumulation of the stock of imported machinery and equipment that
increases the total effective capital.
These simple reduced-form estimations provide us with some reasonable results.
Corresponding to the econometrical analysis in the previous sections, the reduced-form
estimation model confirms that international trade (both export and import) generally plays
emission-increasing role through scale effect but emission decreasing role via composition
4
Agras and Chapman (1999) did the same arrangement in their paper.
Due to data unavailability, we use the stock of imported manufacture goods as an approximation for the stock of
imported equipment and machinery goods. This data series is compiled by author according to the provincial
level statistical report in Almanac of China's Foreign Economic Relationship and Trade (1984-2001). We
consider the stock of imported manufacture goods as a good approximation of the stock of imported equipments
and machinery due to two reasons. Firstly, many of the imported manufacture goods in China are desitinated to
re-export after being assembled. This type of imports is obviously helpful in increasing the productivity of
Chinese industries. Secondly, the manufactured goods importation and the equipment and machinery importation
data on the national level data are available since 1985, comparing these two series of data shows that the imports
of the equipment and machinery occupied a very important and table percentage in total manufacture good
import (over 50%). A simple correlation test between these two series also test reveals very strong correlation
between these two data series by a coefficient of 0.9985.
5
and technique effects. However, a common point among the three tables of reduced-form
estimation results is the unsatisfactory significance for all the three sets of trade intensity
measures. This actually reminds us the underlying complexity of the trade-emission
relationship.
Table 5-8 report the indirect environmental role of international trade on emission going
through its three structural determinants. In the six columns related to 2SLS structural model,
we separately use the three trade intensity measures as instruments for scale effect. The first-
stage instrumentation estimation results are reported in the bottom of the table.
Table 2. The impact of international trade in the decomposed Scale effect (Reduced-form)
Model A
Model B
Model C
(X+M)/GDP (%)
X/GDP (%)
M/GDP (%)
FE
3.293
(1.66)*
RE
FE
RE
FE
4.746
(1.48)°
0.061
(0.02)
5.631
(1.53)°
1.650
(0.72)
3.214
(1.27)
5.962
(2.00)**
0.019
0.019
(4.47)***
(6.45)***
188.343
190.313
195.788
196.876
195.887
202.453
y1993
(4.95)***
(3.36)***
(5.01)***
(3.42)***
(5.27)***
(3.54)***
160.962
153.414
163.272
154.469
173.741
167.072
y1994
(4.31)***
(3.35)***
(4.35)***
(3.38)***
(4.86)***
(3.70)***
129.408
124.570
128.835
123.501
139.096
133.672
y1995
(3.63)***
(3.12)***
(3.60)***
(3.10)***
(4.03)***
(3.46)***
124.511
124.939
126.406
126.600
132.545
134.254
y1996
(3.59)***
(3.52)***
(3.63)***
(3.56)***
(3.94)***
(3.89)***
110.482
112.266
105.908
108.594
112.797
112.169
y1997
(3.18)***
(3.29)***
(3.01)***
(3.19)***
(3.36)***
(3.49)***
113.933
118.420
109.917
115.547
114.424
117.031
y1998
(3.20)***
(3.08)***
(3.05)***
(3.00)***
(3.35)***
(3.30)***
34.776
38.628
33.196
37.749
36.895
40.237
y1999
(0.95)
(1.06)
(0.90)
(1.04)
(1.05)
(1.22)
30.386
29.736
29.208
28.733
34.090
32.774
y2000
(0.80)
(0.72)
(0.77)
(0.69)
(0.93)
(0.88)
66.121
65.924
66.160
65.933
60.307
60.001
trend
(14.87)***
(9.07)***
(14.83)***
(9.11)***
(13.42)***
(8.53)***
-45.802
-53.572
-31.907
Constant
(0.53)
(0.63)
(0.39)
0.13
0.59
0.14
0.59
0.16
0.62
R-squared
964.31
915.19
928.66
Breusch-Pagan
(0.000)
(0.000)
(0.000)
1.67
24.36
2.73
Hausman
(0.9983)
(0.0113)
(0.9938)
290
290
290
290
290
290
Observations
29
29
29
29
29
29
Number of province
Absolute value of z statistics in parentheses, ° significant at 15%, * significant 10%, ** significant 5%, *** significant 1%.
Mstock
RE
2.155
(1.64)*
Table 3. The impact of international trade in the decomposed composition effect (reduced-form)6
Model A
(X+M)/GDP (%)
RE
0.194
(0.64)
Model B
FE
0.793
(1.60)
X/GDP (%)
M/GDP (%)
Model C
RE
FE
RE
FE
2.172
(2.52)**
-1.616
(2.11)**
2.507
(2.86)***
-0.421
(0.53)
0.770
(1.24)
2.227
(2.61)***
Mstock
y1993
y1994
y1995
y1996
y1997
y1998
y1999
y2000
trend
Constant
Observations
Number of province
R-squared
Breusch-Pagan
Hausman
22.811
23.848
(1.86)*
(1.65)°
17.461
13.489
(1.47)
(1.04)
74.460
71.914
(6.50)***
(5.09)***
53.440
53.665
(4.77)***
(5.45)***
32.589
33.528
(2.91)***
(3.66)***
22.867
25.227
(2.00)**
(2.56)**
-9.964
-7.937
(0.85)
(0.71)
2.748
2.406
(0.22)
(0.18)
14.405
14.301
(10.04)***
(8.15)***
-36.491
(2.12)**
290
290
29
29
0.21
0.45
467.23
(0.000)
2.20
(0.9945)
28.629
28.667
(2.29)**
(2.02)**
20.100
14.299
(1.68)°
(1.13)
74.491
71.149
(6.49)***
(5.04)***
54.928
54.885
(4.88)***
(5.67)***
28.709
30.819
(2.53)**
(3.33)***
19.161
23.094
(1.66)°
(2.32)**
-11.636
-8.601
(0.98)
(0.77)
1.783
1.593
(0.15)
(0.12)
14.456
14.309
(10.05)***
(8.29)***
-39.714
(2.44)**
290
290
29
29
0.29
0.46
353.40
(0.000)
20.91
(0.0343)
0.001
0.001
(0.50)
(0.74)
24.431
27.896
(1.98)**
(1.95)*
17.078
13.516
(1.44)
(1.08)
73.984
71.080
(6.44)***
(5.04)***
54.064
54.941
(4.82)***
(5.61)***
32.315
31.955
(2.88)***
(3.54)***
23.022
24.379
(2.02)**
(2.55)**
-9.508
-7.759
(0.81)
(0.71)
2.617
1.903
(0.21)
(0.14)
14.166
14.029
(9.46)***
(7.80)***
-41.714
(2.42)**
290
290
29
29
0.22
0.46
439.96
(0.000)
6.99
(0.7996)
6
For the aim of correspondence to the empirical analyses in chapter 4, in the appendix of this chapter, we report
both the reduced-form and 2SLS estimation results for the decomposed composition effect without time trend and
time-specific effects.
Table 4. The impact of international trade in the decomposed technique effect (reduced-form)
(X+M)/GDP (%)
Model A
RE
FE
-2.594
-4.129
(2.17)**
(2.26)**
X/GDP (%)
M/GDP (%)
Model B
Model C
RE
FE
RE
FE
-6.141
(2.08)**
0.322
(0.13)
-7.750
(2.38)**
-1.564
(0.73)
-4.792
(2.05)**
-8.261
(2.74)***
Mstock
y1993
y1994
y1995
y1996
y1997
y1998
y1999
y2000
trend
Constant
Observations
Number of provcode
R-squared
Breusch-Pagan
Hausman
-205.196
-207.855
(5.81)***
(3.82)***
-163.827
-153.643
(4.72)***
(3.43)***
-178.305
-171.777
(5.38)***
(4.54)***
-178.315
-178.892
(5.53)***
(5.18)***
-151.691
-154.099
(4.70)***
(4.72)***
-75.298
-81.352
(2.28)**
(2.63)***
-20.531
-25.729
(0.60)
(0.80)
5.934
6.812
(0.17)
(0.17)
-70.189
-69.923
(16.99)***
(10.12)***
105.935
(1.39)
290
290
29
29
0.15
0.64
927.80
(0.000)
4.08
(0.9437)
-215.421
-218.031
(5.95)***
(3.99)***
-167.189
-155.367
(4.80)***
(3.50)***
-177.626
-170.170
(5.35)***
(4.51)***
-180.918
-181.468
(5.59)***
(5.30)***
-145.340
-148.373
(4.45)***
(4.50)***
-69.654
-76.838
(2.09)**
(2.47)**
-18.262
-24.320
(0.54)
(0.76)
7.564
8.410
(0.21)
(0.21)
-70.247
-69.940
(16.97)***
(10.19)***
115.963
(1.54)
290
290
29
29
0.17
0.65
873.65
(0.000)
8.19
(0.6961)
-0.012
-0.012
(3.01)***
(4.85)***
-214.369
-222.667
(6.09)***
(4.08)***
-173.280
-164.880
(5.11)***
(3.73)***
-184.125
-177.299
(5.63)***
(4.78)***
-184.528
-186.704
(5.79)***
(5.50)***
-150.506
-149.732
(4.73)***
(4.73)***
-73.276
-76.584
(2.26)**
(2.64)***
-20.979
-25.214
(0.63)
(0.83)
4.292
5.941
(0.12)
(0.16)
-66.510
-66.107
(15.62)***
(9.63)***
101.800
(1.38)
290
290
29
29
0.17
0.65
875.98
(0.000)
4.44
(0.9551)
Table 5. Trade’s impact on industrial SO2 emission—Channel 1: enlargement of the
industrial production scale
Simple
Ind. GDP
y1993
y1994
y1995
y1996
y1997
y1998
y1999
y2000
trend
Constant
Obs.
Group #
R-squared
BreuschPagan
Hausman
Iden. test
RE
0.481
(16.69)***
186.420
(6.87)***
176.830
(6.80)***
144.485
(5.73)***
132.457
(5.34)***
116.792
(4.71)***
117.905
(4.67)***
42.062
(1.62)°
40.271
(1.48)°
28.169
(7.19)***
-103.628
(1.93)*
290
29
FE
0.480
(8.52)***
186.415
(4.36)***
176.825
(4.94)***
144.467
(4.77)***
132.429
(4.99)***
116.762
(4.66)***
117.865
(4.34)***
42.016
(1.85)*
40.244
(1.73)*
28.290
(5.94)***
-103.267
(1.84)*
290
29
0.79
934.48
(0.000)
Model A
2SLS
Model B
RE
0.421
(2.59)***
186.192
(6.99)***
176.632
(6.92)***
143.740
(5.79)***
131.353
(5.36)***
115.570
(4.71)***
116.332
(4.64)***
40.223
(1.55)°
39.180
(1.46)
33.003
(2.48)**
-89.187
(1.07)
290
29
0.60
FE
0.595
(1.66)*
186.848
(3.32)***
177.201
(3.68)***
145.882
(3.48)***
134.526
(3.62)***
119.082
(3.37)***
120.851
(3.10)***
45.508
(1.21)
42.316
(0.99)
19.115
(0.68)
-130.678
(1.12)
290
29
0.60
RE
0.268
(2.34)**
185.619
(6.34)***
176.134
(6.28)***
141.868
(5.21)***
128.580
(4.79)***
112.502
(4.19)***
112.384
(4.11)***
35.606
(1.26)
36.441
(1.24)
45.137
(4.64)***
-52.938
(0.74)
290
29
0.56
FE
0.350
(1.72)*
185.926
(3.33)***
176.401
(3.69)***
142.870
(3.48)***
130.065
(3.63)***
114.145
(3.32)***
114.498
(3.05)***
38.079
(1.06)
37.908
(0.90)
38.639
(2.33)**
-72.350
(0.84)
290
29
0.59
Model C
RE
FE
0.280
0.296
(6.13)***
(6.22)***
185.663
185.725
(6.36)***
(3.34)***
176.173
176.226
(6.30)***
(3.80)***
142.012
142.213
(5.24)***
(3.59)***
128.794
129.091
(4.83)***
(3.77)***
112.738
113.067
(4.23)***
(3.50)***
112.687
113.111
(4.15)***
(3.21)***
35.961
36.457
(1.29)
(1.12)
36.651
36.945
(1.26)
(0.99)
44.204
42.901
(8.87)***
(5.66)***
-55.724
(0.89)
290
290
29
29
0.56
0.57
0.014
(1.000)
0.030
(1.000)
3.70
(0.9599)
0.94
(0.9999)
32.24
26.10
(0.0003)*** (0.0036)***
0.02
(1.000)
Trade elasticity of decomposed emission changes caused by industrial production scale enlargement
0.152
0.193
(X+M)/GDP
0.164
0.194
X/GDP
0.104
-0.035
-0.039
M/GDP
Mstock
0.060
First-stage coefficient for the international trade instruments (on ind. GDPjt)
(X+M)/GDP
(%)
X/GDP (%)
M/GDP (%)
6.160
(3.05)***
0.123
0.064
5.534
(1.36)
20.580
(4.48)***
-4.536
(1.21)
18.669
(2.04)**
-3.808
(0.68)
11.897
(4.20)***
14.056
(3.23)***
0.067
0.067
(14.41)*** (16.14)***
d(scale effect )
The detailed calculation equation for elasticity of export intensity ratio for decomposed scale effect is as following,
d(X / GDP) =
Mstock
14.056× X / GDP ×0.296×ind .GDP = 14.096×0.296× X / GDP=0.123 . Here 14.096 and 0.296 are the coefficients obtained for
ind .GDP
SO2
SO2
export intensity in the instrumentation estimation and for industrial GDP in the second-stage fixed-effect estimation result of model C,
respectively. The X / GDP, ind.GDP and SO2 mean the sample average value of X/GDP, industrial GDP and the decomposed industrial SO2
emission variation related to the scale effect. The elasticity of trade for the other two decomposed emission changes is calculated in the same
way.
Table 6. Trade’s impact on industrial SO2 emission—Channel 2: transformation of
industrial composition
Simple
(K/L)jt
(1/10000)
y1993
y1994
y1995
y1996
y1997
y1998
y1999
y2000
trend
Constant
Obs.
Group #
R-squared
BreuschPagan
Hausman
Overiden.
RE
FE
-6.268
-7.591
(4.15)***
(4.30)***
20.722
20.352
(1.73)*
(1.47)
13.909
12.888
(1.21)
(1.06)
68.721
67.335
(6.11)***
(4.66)***
45.804
44.207
(4.12)***
(4.41)***
24.019
22.273
(2.16)**
(2.43)**
19.275
18.678
(1.73)°
(2.06)**
-12.528
-12.931
(1.09)
(1.24)
2.447
2.360
(0.20)
(0.18)
17.848
18.568
(10.98)***
(9.92)***
-11.073
(0.73)
290
290
29
29
0.32
0.47
412.71
(0.000)
0.74
(1.000)
2SLS
Model B
Model A
RE
1.145
(0.14)
22.796
(1.68)°
19.630
(1.38)
76.483
(5.11)***
54.749
(3.52)***
33.796
(2.10)**
22.619
(1.75)*
-10.272
(0.79)
2.934
(0.22)
13.815
(3.01)***
-35.108
(1.21)
290
29
0.18
FE
-47.153
(1.60)°
9.283
(0.56)
-17.644
(0.68)
25.906
(0.77)
-3.530
(0.10)
-29.905
(0.76)
0.832
(0.05)
-24.974
(1.86)*
-0.243
(0.02)
40.090
(2.47)**
FE
-15.091
(0.63)
18.253
(1.15)
7.100
(0.32)
59.482
(2.06)**
35.158
(1.17)
12.382
(0.38)
15.295
(1.12)
-15.214
(1.23)
1.866
(0.14)
22.648
(1.71)°
290
29
0.64
RE
-12.310
(1.82)*
19.031
(1.53)
9.246
(0.72)
62.393
(4.63)***
38.513
(2.77)***
16.049
(1.12)
16.549
(1.40)
-14.368
(1.20)
2.049
(0.17)
21.135
(5.34)***
8.518
(0.32)
290
29
0.27
0,91
(0,999)
0,74
(1,000)
0,84
(0,999)
0,08
(1,000)
290
29
0.64
Model C
RE
FE
7.923
-51.638
(0.28)
(2.16)**
24.692
8.028
(1.55)
(0.51)
24.861
-21.105
(0.98)
(0.94)
83.581
21.211
(2.62)***
(0.74)
62.927
-8.941
(1.75)°
(0.30)
42.734
-35.819
(1.10)
(1.12)
25.676
-1.191
(1.42)
(0.09)
-8.209
-26.339
(0.52)
(2.11)**
3.380
-0.538
(0.24)
(0.04)
10.128
42.530
(0.67)
(3.30)***
-57.082
(0.62)
290
290
29
29
0.01
0.64
0,23
(1,000)
Trade elasticity of decomposed emission changes caused by industrial composition transformation
(X+M)/GDP
0.013
0.314
X/GDP
0.208
-0.003
0.030
M/GDP
-0.269
0.088
Mstock
-0.008
First-stage coefficient for the international trade instruments (on (K/L)jt)
(X+M)/GDP
(%)
X/GDP (%)
M/GDP (%)
0.029
(2.81)**
0,79
(0,999)
0.164
0.041
-0.017
(0.79)
-0.085
(2.59)***
0.113
(3.89)***
0.001
(0.03)
-0.030
(0.73)
0.019
(0.82)
-0.016
(0.55)
-0.00005
-0.00004
(0.84)
(0.95)
maccunew
Table 7. Trade’s impact on industrial SO2 emission—Channel 2: transformation of
industrial composition (New composition indicator)
Simple
Comp.
indicator
y1993
y1994
y1995
y1996
y1997
y1998
y1999
y2000
trend
Constant
Obs.
Group #
R-squared
BreuschPagan
Hausman
Overiden.
RE
FE
7.664
8.556
(7.80)***
(5.67)***
17.157
16.538
(1.56)
(1.22)
16.180
15.882
(1.53)
(1.35)
49.416
46.405
(4.60)***
(3.98)***
33.607
31.307
(3.24)***
(3.49)***
18.268
16.636
(1.79)*
(2.11)**
7.632
5.948
(0.73)
(0.69)
-26.172
-27.982
(2.44)**
(2.91)***
2.739
2.725
(0.25)
(0.25)
7.748
6.969
(5.01)***
(4.15)***
-166.438
(7.33)***
290
290
29
29
0.22
0.56
528.05
(0.000)
6.82
(0.742)
2SLS
Model B
Model A
RE
0.834
(0.22)
21.897
(1.48)
18.467
(1.32)
72.469
(3.91)***
51.217
(3.12)***
30.760
(2.06)**
20.527
(1.35)
-12.313
(0.78)
2.846
(0.20)
13.710
(3.71)***
-46.092
(0.68)
290
29
0.24
FE
-31.842
(1.60)°
44.574
(2.21)**
29.407
(2.06)**
182.758
(2.65)***
135.464
(2.54)**
90.521
(2.35)**
82.222
(2.07)*
53.990
(1.26)
3.358
(0.25)
42.233
(2.40)**
FE
7.706
(1.08)
17.128
(1.11)
16.166
(1.25)
49.276
(1.77)*
33.500
(1.59)
18.192
(1.11)
7.554
(0.44)
-26.256
(1.38)
2.738
(0.20)
7.712
(1.19)
290
29
0.55
RE
4.954
(0.85)
19.038
(1.61)°
17.088
(1.58)
58.563
(2.65)***
40.595
(2.25)**
23.225
(1.58)
12.749
(0.85)
-20.673
(1.30)
2.781
(0.25)
10.114
(1.94)
-118.685
(1.15)
290
29
0.25
3,53
(0,9661)
0,68
(1,000)
0,23
(1,000)
0,01
(1,000)
290
29
0.56
Model C
RE
FE
-6.421
-1.187
(0.97)
(0.20)
26.932
23.299
(1.75)*
(1.52)
20.896
19.144
(1.47)
(1.49)
96.955
79.290
(3.72)***
(3.23)***
69.921
56.427
(3.23)***
(3.04)***
44.028
34.455
(2.44)**
(2.36)**
34.225
24.343
(1.86)*
(1.61)°
2.408
-8.213
(0.12)
(0.50)
2.960
2.878
(0.20)
(0.21)
20.043
15.474
(3.34)***
(2.92)***
81.740
(0.70)
290
290
29
29
0.07
0.57
4,67
(0,9119)
Trade elasticity of decomposed emission changes caused by industrial composition transformation
(X+M)/GDP
-0.015
0.308
X/GDP
0.047
0.138
0.029
M/GDP
-0.110
0.158
Mstock
0.025
First-stage coefficient for the international trade instruments (on new composition indicator)
(X+M)/GDP
(%)
X/GDP (%)
M/GDP (%)
-0.047
(4.52)***
2,18
(0,9948)
-0.008
0.0005
-0.025
(0.69)
0.048
(0.96)
-0.113
(2.62)***
0.090
(1.42)
0.106
(1.94)*
-0.023
(0.693)
0.034
(0.72)
-0.0002
-0.0002
(2.87)***
(3.40)***
Detailed introduction for the new composition indicator is available in appendix of this chapter.
maccunew
Table 8. Trade’s impact on industrial SO2 emission—Channel 3: technique progress
Simple
RE
(GDPPC)jt
(1/1000)
(GDPPC)jt2
(1/1000)2
(GDPPC)jt3
(1/1000)3
y1993
y1994
y1995
y1996
y1997
y1998
y1999
y2000
trend
Constant
Obs.
Group #
R-squared
BreuschPagan
Hausman
Overiden.
2SLS
Model B
Model A
FE
-248.380
-438.174
(3.64)***
(3.15)***
19.680
36.014
(2.59)***
(2.91)***
-0.547
-1.069
(1.90)*
(2.69)***
-190.633
-183.624
(5.56)***
(3.74)***
-163.845
-152.055
(4.95)***
(3.71)***
-169.593
-156.267
(5.27)***
(4.58)***
-155.992
-141.582
(4.92)***
(4.52)***
-126.321
-111.637
(3.98)***
(3.68)***
-46.997
-34.025
(1.46)
(1.21)
-3.015
2.166
(0.09)
(0.07)
5.560
4.183
(0.16)
(0.11)
-35.531
-9.297
(4.01)***
(0.54)
446.931
(3.53)***
290
290
29
29
0.12
0.67
891.11
(0.000)
13.85
(0.3103)
Model C
RE
FE
RE
FE
RE
FE
-56.935
(2.04)**
-121.885
(2.26)**
-62.193
(2.25)**
-125.118
(2.39)**
-81.423
(3.70)***
-103.787
(4.43)***
-201.451
(5.58)***
-181.896
(5.26)***
-190.346
(5.68)***
-178.013
(5.40)***
-147.182
(4.46)***
-63.941
(1.91)*
-15.181
(0.44)
-4.796
(0.13)
-51.677
(5.07)***
148.596
(1.74)*
290
29
0.13
-202.302
(3.75)***
-182.881
(3.94)***
-191.499
(4.94)***
-178.783
(5.18)***
-146.678
(4.57)***
-62.654
(2.13)**
-19.096
(0.60)
-15.345
(0.38)
-30.047
(1.71)*
-202.344
(3.78)***
-182.930
(3.94)***
-191.557
(4.96)***
-178.821
(5.23)***
-146.653
(4.58)***
-62.589
(2.14)**
-19.291
(0.61)
-15.870
(0.39)
-28.970
(1.69)°
290
29
0.73
-201.771
(5.48)***
-182.267
(5.17)***
-190.781
(5.58)***
-178.303
(5.30)***
-146.992
(4.37)***
-63.456
(1.86)*
-16.657
(0.47)
-8.773
(0.24)
-43.522
(5.12)***
196.278
(2.39)**
290
29
0.09
-202.065
(3.79)***
-182.607
(4.05)***
-191.178
(5.06)***
-178.568
(5.30)***
-146.819
(4.63)***
-63.012
(2.19)**
-18.005
(0.59)
-12.406
(0.32)
-36.074
(3.79)***
290
29
0.73
-201.519
(5.56)***
-181.976
(5.24)***
-190.439
(5.66)***
-178.076
(5.38)***
-147.141
(4.44)***
-63.837
(1.90)*
-15.498
(0.45)
-5.650
(0.15)
-49.926
(4.93)***
158.835
(1.86)*
290
29
0.12
1,59
(0,9986)
3,94
(0,9501)
2,21
(0,9945)
4,77
(0,9057)
9,33
(0,5014)
10,65
(0,3853)
-0.173
-0.205
-0.025
-0.032
0.069
(7.04)***
0.064
(3.74)***
Trade elasticity of decomposed emission changes caused by technical progress
(X+M)/PIB
-0.145
-0.253
X/GDP
-0.117
-0.228
M/GDP
-0.051
-0.060
Mstock
First-stage coefficient for the international trade instruments (on GDPPCjt)
(X+M)/GDP
(%)
X/GDP (%)
M/GDP (%)
0.042
(8.06)***
290
29
0.79
0.034
(3.09)***
0.061
(4.61)***
0.027
(2.48)**
0.059
(3.04)***
0.016
(1.20)
0.0001
0.0001
(7.87)***
(4.20)***
Mstock
Let’s first look at the results concerning the environmental role of trade through
production scale enlargement impact in table 5. Good coherence between the results of the
second-stage 2SLS estimation and those in the reduced-form model can be easily observed in
this table. In two to three cases, the changes in the trade intensity measurement and estimation
methods do not affect the coefficients of the trade variables. The only exception is the model
B, where we only prove the significant positive coefficients for export intensity. As the
Hausman specification test indicates the best instrumentation efficiency to be in Model C, in
which we use X/GDP and Mstock/GDP to instrument industrial GDP. We believe this result
actually confirms the assumption of the CGE model of de Melo and Robinson (1990) on its
production function—the growth of domestic economy benefits the positive spillover effect
from both the export intensity increase and the accumulation in the imported equipment and
machinery. Therefore, we will base our following discussion on the result of model C. To sum
up, the estimation results in both tables 2 and 5 generally confirm the positive causality
between the trade intensity (either total trade intensity or annual export intensity and the stock
of imported machinery and equipment), the industrial production scale, and the decomposed
industrial emission variation related to scale effect. This conclusion is also expressed in the
table 5 by the positive trade elasticity of the decomposed scale effect.
The estimations for the decomposed composition effect do not obtain such a good
coherence. Most of the coefficients for the composition effect measure—capital/labor
abundance ratio shows negative coefficients even after being instrumented by the trade
intensity variables (c.f. fixed effect estimation results in table 6). This finding is actually
contradictory with our intuition that suggests an industrial sectors using intensively capital in
its production process generate equally more intensively pollution problems.
This contrary-to-intuition result may be due to the fact that the capital-abundance ratio,
being a pertinent extrapolation for the composition effect when analysis focus covers the
whole economy, may lose its pertinence when the analysis is concentrated only in the
industrial sector. Here is our explanation. The pollution performance of whole economy
structure and the total capital-abundance level may follow similar evolution trajectory when
economy starts from less capital-intensive and less pollution-intensive agriculture-dominated
structure, then experiences the industrialization procedures that leads both capital intensity
and pollution intensity to increase and finally arrives the post-industry structure whose
production process returns to be low capital-intensive and low pollution intensive style.
However, once our attention is focused only in the industrial sectors, less polluting industries
might not necessarily be the sectors possessing less capital abundance ratio, the example is the
information industries. Moreover, in the relatively early stage of industrialization procedure,
though the less polluting light industries may generally possess relatively lower capital
abundance ratio, for the enterprises in the same industry, higher capital abundance ratio may
means better technology efficiency, therefore means less pollution problems. Dinda et al
(2000) also indicate the potential ambiguity in using capital abundance as measurement for
environmental performance of the industrial composition, since the “capital intensive sector
could also be more likely to be clean technology owner”.
To overcome the ambiguity of this simple composition effect measurement, we make
use of the detailed value added data of 13 industrial sectors in each province to construct a
Y jit
synthetic composition indicator for each provinces as following, Ωit = ∑( Yit × I j,0 ) . Yjit signifies
j
the detailed value added in industrial sector j of province i at time t and Ij,0 is the initial
national average SO2 emission intensity for each of the 13 sectors in year 1991.7 As the sum of
value added of the 13 industrial sectors generally counts up to 98% of the total provincial
industrial GDP each year, we are confident about the representative capability of this synthetic
indicator as the measurement for the general environmental performance of the industrial
composition.8
The 2SLS structural model estimation results based on this new composition effect
measurement are listed in table 7. The new synthetic composition indicator does improve the
reliability of the estimation results. The expected positive coefficients are firstly obtained by
the simple estimations. However, for the structural models, most of these positive coefficients
stay insignificant and the trade intensity variables also show low instrumentation efficiency.
We therefore fail to capture significant impact of trade on emission through China’s industrial
composition transformation.
The estimation results obtained for the decomposed emission variations related to
technique effect are relatively more satisfactory (c.f. Table 8). As expected, per capita GDP is
a pertinent approximation for the technique effect, especially when collectively instrumented
by the trade intensity indicators X/GDP and Mstock. In this estimation result, the positive
GDPPC coefficient shows the highest significance. The estimation proves an obvious
environment-friendly role for international trade through technique effect. This conclusion is
also revealed in Table 8 by the unanimous negative trade elasticity.
7
More discussion about the advantages and deficiencies of this synthetic composition effect measurement is
included in the appendix.
8
Keller and Levinson (2002) also use the same expression to measure industry composition of each state in order
to adjust the measurement for state pollution abatement costs.
6. Conclusion
In this paper, we introduced and adjusted the traditional Divisia index decomposition
method and employed it to Chinese data. By deepening our analysis insight into the detailed
13 industrial sectors in each province, the Divisia decomposition method revealed more
concrete contribution of industrial scale enlargement, composition transformation and
technical progress in industrial SO2 emission changes for each province during 1991-2001.
The decomposition results reveals the contributions of the three structural determinants
of emission in China’s industrial SO2 emission variation, it gives more structural explanation
for the ever-increasing tendency in total industrial SO2 emission in China during the last
decade. Besides the contribution from the rapid expansion in industrial production scale, we
find that the industrial composition transformation process also plays as pollution-increasing
factor in most of Chinese provinces. Only several coastal provinces whose industrialization
largely benefited from their advantageous geographical location and openness policy have
been the exceptions. Accompanying scale enlargement, the economic growth in each province
also catalyzed the technique effect, a pollution-reducing factor. As expected, it is the ricer
coastal provinces possess a technique effect relativement more important owing to their
relatively higher income level.
Coherent to most of previous emission decomposition studies, our findings show the
scale and technique effects as the two most important emission variation contributors, but in
opposite directions. This actually leads the final changes in SO2 emission to be determined by
the combination of the composition effect with the left impact from the mutual cancellation
between scale and technique effect. However, when the analysis is précised to the special case
of each province, the variation of its emission during the time is in fact a particular result,
which is determined by the combination of its three specific structural characteristics.
The Divisia decomposition results also permit us to separately investigate the potential
relationship between the trade intensity indicators and the three parts of emission variation
caused scale enlargement, composition transformation and technical progress through both a
reduced-form model and a structural model estimated by 2SLS method. This simple
econometrical strategy clarified the relationship of trade with both scale and technique effects.
We prove the significant role of the trade as a positive factor in production scale enlargement,
and an active catalyst for the technological progress in pollution control activities but failed to
derive a clear conclusion for the role of trade on China’s industrial composition
transformation.
Reference
Agras, J. and D. Chapman (1999), ‘A dynamic approach to the Environmental Kuznets Curve hypothesis’.
Ecological Economics 28(2): 267-277.
Ang, B.W., Pandiyan, G., 1997. Decomposition of energy-induced CO2 emissions in manufacturing, Energy
Economics 19, 363-374.
Baltagi, B.H. (1995), Econometric Analysis of Panel Data, Wiley and Sons, Ltd, West Sussex, England, 304pp.
Bruvoll, A. and H. Medin, (2003). Factors Behind the Environmental Kuznets Curve: A Decomposition of the
Changes in Air pollution. Environmental and Resource Economics, 24(2003):27-48.
Choi, K.H., B.W. Ang, (2001). A time-series analysis of energy-related carbon emission in Korea. Energy Policy,
29(2001), 1155-1161.
Choi, K.H., B.W. Ang, (2002). Measuring thermal efficiency improvement in power generation : the Divisia
decomposition approach. Energy, 27(2002), 447-455.
Choi, K.H., B.W. Ang, and K.K.Ro, (1995). Decomposition of the energy-intensity index with application for the
Korean Manufacturing Industry. Energy, 20(9), 835-884.
De Melo, J., and S. Robinson (1990), ‘Productivity and Externalities: Models of Export-led Growth’, Memo.
90.10, University of Geneva, Geneva.
Divisia, F. (1925). L’indice monétaire et la théorie de la monnaie. Revue D’Economie Politique, 39(4), 842-864.
Divisia, F. (1926). L’indice monétaire et la théorie de la monnaie. Revue D’Economie Politique, 39(6), 9801008.
Greene, W. H. (2000), Econometric Analysis, International Edition, (Version IV), New York University:
Prentice-Hall International Inc.
Greening, L.A., W. B. Davis, L. Schipper and M. Khrushch (1997). Comparison of six decomposition methods:
application to aggregate energy intensity for manufacturing in 10 OECD countries. Energy Economics, 19
(1997), 375-390.
Grossman, G. (1995). Pollution and growth: What do we know? Goldin and Winter (eds.). The Economics of
Sustainable Development, Cambridge University Press, 1995.
He, J. (2005). Estimation on Economic Cost of China's New De-sulfur Policy during Her Gradual Accession to
WTO: The Case of Industrial SO2 Emission. China Economic Review, forthcoming.
Keller, W. and A. Levinson, (1999). ‘Pollution Abatement Costs and Foreign Direct Investment Inflow to U.S.
States’. NBER. Working Paper no. 7369, Cambridge, MA.
Keller, W. and A. Levinson, (2001). ‘Pollution Abatement Costs and Foreign Direct Investment Inflow to U.S.
States’. The Review of Economics and Statistics,LXXXIV (4): 691-703.
Lawrence Berkeley National Laborarory, 2001. China Energy Databook,.5.0, Berkeley , 2001.
OECD (1991)??
SEPA(State Environmental Protection Agency, PRC) (1986-1997), China Environmental Yearbook. Beijing:
China Statistical Publishing House.
SEPA(State Environmental Protection Agency, PRC) (1998-2002), China Environmental Statistic Yearbook.
Beijing.
SSB(State Statistical Bureau, PRC) (1984-2004), China Statistical Yearbook, Beijing: China Statistical
Publishing House.
SSB(State Statistical Bureau, PRC) (1986-1997), China Industrial Economics Statistics Yearbook. Beijing:
China Statistical Publishing House.
SSB(State Statistical Bureau, PRC) (1989-2004). China Statistical Yearbook, China Statistical Publishing house,
Beijing.
Suri, V. and D. Chapman, (1998). ‘Economic growth, trade and energy: implications for the environmental
Kuznets curve’, Ecological Economics, 25(2): 195-208.
Törnqvist, L. (1936). The Bank of Finland’s Consumption price index. Bank of Finland Monthly Bulletin, 10,
168.
Torvanger, A. (1991). Manufacturing sector carbon dioxide emissions in nine OECD countries, 1973–87 : A
Divisia index decomposition to changes in fuel mix, emission coefficients, industry structure, energy
intensities and international structure. Energy Economics, 13(3) :168-186.
Appendix. A detailed introduction on the synthetic composition indicator
The original idea for the construction of this synthetic composition indicator is due to
the unsatisfactory estimation results of the composition indicator—(K/L), the capital
abundance ratio in industrial sector. We believe a efficient composition measurement must
include both the detailed production structure and the emission performance of each industrial
sector that actually exist in industrial economy.
Inspired by the Grossman decomposition, we have,
Y j,it
SO2,it Yit
SO2,it
=
× ∑(
× e j,it )×
areai 
area
i
it
 j Yit
Y


 Technique .
scale 
Compositio n
where ej,it is the emission efficiency indicator for each sector j of province i, we define
the synthetic composition indicator by fixing the emission efficiency indicator ej,it to equal to
the Ij0—the national level average emission intensity of each industrial sector j at base year
1990, so ej,it=Ij,1990. Therefore, for a province i in each t, we can construct an indicator as Ωit=
∑YYjit,it ×I j,1990 . In this indicator, as the sector-specific emission intensity is fixed at the initial
j
level, for each province, its variations during the time actually reveals the changes of the ratio
of the different industrial sectors. Therefore, when the ratio of a relatively more polluting
sector experiences an increase during period 1, we can expect an increase in this indicator, so
Ωit>Ωi0. Therefore, we use this indicator as a synthetic composition index.
For each province, this indicator is actually constructed by combining the detailed value
added data of the 13 industrial sectors of each province in different years with the sectorspecific emission intensity in base year. The 13 industrial sectors are total mining industry,
food and beverage, textile, paper, total power industry, chemical materials, pharmacy, fiber,
non-metal products, metal processing and smelting, metal products, machinery and the other
industry, these data comes from China Industrial Economic Statistic Yearbook (1989-2002),
and their total value added accounts up to 98% of the total industrial VA in each province.
Therefore we are quite confident on the representation capacity of this indicator.
We are, however, not the first authors using this indicator in emission-related analysis.
Keller and Levinson (2002) also use the same expression to measure industry composition of
each state in order to adjust the measurement for state pollution abatement costs.

WHAT IS THE ROLE OF OPENNESS FOR CHINA’S ENVIRONMENT? AN

Transcription

Similar documents

How to reduce emissions most effectively? COHIBA Final Conference October 11-12

Electromagnetic Compatibility of Integrated circuits IC-EMC

Document 6516450

Contents of Galaxies: The Interstellar Medium

SMP alumni e-news - Crcna - Christian Reformed Church

Mas d`intras - Verity Wines

More Clean, More Green

February 2010 - Utah Environmental Health Association

SAVECO

Managed EN 50155 Switch